The long-term objective of this project is to develop statistical methods and computer programs for effective analysis of microarray data. The development will build on our previous work on the model-based approach to such analyses but will address several issues in order to significantly improve the performance. Our specific aims are as follows. Aim 1. To develop methods for the full modeling of the responses of perfect match and mismatch oligonucleotide probes to changes in mRNA concentration, and to exploit such models to construct estimates of improved expression indexes. By doing so, we hope to extend the linearity of the estimates to the low signal range where cross-hybridization effects must be accounted for in the estimation process. We will also investigate the feasibility of oligonucleotide expression arrays based on perfect match-only design. Aim 2. To develop array-to-array normalization methods applicable to situations when the corresponding samples are expected to exhibit very different expression patterns. Aim 3. To develop statistical methods for the analysis of tag-probe arrays for the study of mixtures of yeast mutants. Aim 4. To develop computer program modules based on the above methods, and to incorporate them into the software dChip in order to extend its functionality for model-based analysis of oligonucleotide array data. We will also work closely with experimental groups in the analysis of real data, in order to ensure that the results from our research will be relevant to the need of the user community. We will aim to improve dChip not only by the methods directly resulting from Aim 1, but will also selectively implement and incorporate other useful analysis methods developed by our group and our collaborators into the dChip framework. Aim 5. To collaborate with the R-based bioinformatics consortium to facilitate the transfer of our methodology to that platform, and to extend the functionality of dChip through R function calls.